def embed_vid(outvid):
    video = io.open(outvid, 'r+b').read()
    encoded = base64.b64encode(video)
    return HTML(data='''<video alt="test" width="950" height="500" loop="true" controls>
                <source src="data:video/mp4;base64,{0}" type="video/mp4" loop="true" />
             </video>'''.format(encoded.decode('ascii')))

Storm Track Analysis

HTML(showcode)

The tracking algorithm is a fork of the Tint (Tint is not Titan) tracking algorithm (http://openradarscience.org/TINT/). The framework has been modified that it can be also applied to model output data that is not not stored in Py-ART radar data container.

• The analysis is based averages of one week worth of strome cell tracking:

1. Median of strom area, duration, avg., max. rain-rates and # of storm cells

#Plot the Medians
um044_n = len(UM044_ens.coords['dim_0'])
um133_n = len(UM133_ens.coords['dim_0'])
cpol_n = len(OBS.dataset['total'].coords['dim_0'])
cpol_n2 = len(CPOL.coords['dim_0'])
CPOL_pdf2 = ravel(OBS.dataset['obs'])
var=['avg_area','dur', 'avg_mean', 'max_mean', 'dist', 'v']
names=['area', 'duration', 'avg-rain', 'max-rain', 'distance', 'speed', '# storms']
#print(CPOL_pdf[var].median())
medians = pd.DataFrame({'a': list(CPOL_pdf[var].median().values)+[int(cpol_n)],
                        'b': list(CPOL_pdf2[var].median().values)+[int(cpol_n2)],
                        'd': list(UM044_pdf[var].median().values)+[int(um044_n)],
                        'c': list(UM133_pdf[var].median().values)+[int(um133_n)]} )
                       #index=('Area','Duration', 'Mean-Rain', 'Max-Rain'))
medians.index=names
medians.columns=['Bootstrap', 'Cpol', 'UM 1.33km', 'UM 0.44km']
#print('Medians:')
                 
medians.round(2)

	Bootstrap	Cpol	UM 1.33km	UM 0.44km
area	111.20	119.38	81.64	67.34
duration	60.00	74.67	60.00	60.00
avg-rain	4.52	4.50	4.91	5.93
max-rain	6.58	6.60	7.37	9.06
distance	15.10	16.42	10.60	12.09
speed	12.61	12.83	10.03	12.78
# storms	1184.00	42.00	50.00	73.00

2. 2D-Histograms (storm area vs duration):

#Create Hex-Bin plot
mpld3.disable_notebook()
var=['avg_area','dur', 'max_mean', 'avg_mean']
medians = pd.DataFrame({ 'Bootstrap':CPOL_pdf[var].median(),
                        'CPOL':CPOL_pdf2[var].median(),
                        'UM 1.33km':UM133_pdf[var].median(),
                        'UM 0.44km':UM044_pdf[var].median()})
#medians.loc['avg_area'] /= 2.5**2
histdata =  [UM044_pdf[var].dropna(), UM133_pdf[var].dropna(), CPOL_pdf2[var].dropna(), CPOL_pdf[var]][::-1]
titles = ['UM 0.44km ens', 'UM 1.33km ens', 'CPOL', 'Bootstrap'][::-1]
fig = plt.figure(figsize=(10,8))
colm = colmap2.Blues
colm.set_under('w', alpha=0)
fig, (ax1, ax2, ax3, ax4) = plt.subplots(1, 4, sharey=True)
fig.subplots_adjust(bottom=0.07, right=0.98, left=0.01, top=0.35, wspace=0.01)
cbar_ax = fig.add_axes([0.15, 0.02, 0.7, 0.01])
hexbin_data = []
gridsize = 10
vmin=0.0001
vmax=0.1
nticks=10
YMax, XMax = 260, 70*2.5**2
#YMax, XMax = 50, 70*2.5**2
for i, ax in enumerate((ax1, ax2, ax3, ax3)):
    data = histdata[i][var[:2]]
    #data[var[0]] /= 2.5**2
    #data = data.loc[(data[var[0]] <= XMax) & (data[var[1]] <=YMax)]
    X = data[var[0]].values
    Y = data[var[1]].values
    ax.set_ylim(0,YMax)
    ax.set_xlim(0,XMax)
    hb =  ax.hexbin(X, Y,  gridsize=gridsize, extent=(0,XMax,0,YMax))
    hexbin_data.append(np.ones_like(Y, dtype=np.float) / hb.get_array().sum())
    
cl = plt.cla()
medians = OrderedDict()

for i, ax in enumerate((ax1, ax2, ax3, ax4)):
    data = histdata[i][var[:2]]
    #data[var[0]] /= 2.5**2
    #data = data.loc[(data[var[0]] <= XMax) & (data[var[1]] <=YMax)]
    X = data[var[0]].values
    Y = data[var[1]].values
    ax.set_ylim(0,YMax)
    ax.set_xlim(0,XMax)
    im = ax.hexbin(X, Y,  gridsize=gridsize, cmap=colm, marginals=False, extent=(0,XMax,0,YMax),
                     vmin=vmin, vmax=vmax, C=hexbin_data[i], reduce_C_function=np.sum)
    
    ax.set_title(titles[i], fontsize=24)
    ax.grid(color='w', linestyle='', linewidth=0)
    ax.tick_params(labelsize=24)
    ax.xaxis.set_ticks(ax.xaxis.get_ticklocs()[:-1])
    x, y = histdata[i][var[0]].median(), histdata[i][var[1]].median()
    z, zz=  histdata[i][var[2]].median(), histdata[i][var[-1]].median()
    sx, sy = histdata[i][var[0]].std(), histdata[i][var[1]].std()
    medians[titles[i]] = '%2.1f km^2, %2i min (max: %2.1f mm/h, mean: %2.1f mm/h);'%(x,y, z, zz)
    ax.hlines(y*np.ones_like(histdata[i][var[0]]),0,x, 'firebrick', lw=3)
    ax.vlines(x*np.ones_like(histdata[i][var[1]]),0,y, 'firebrick', lw=3)
    ax.scatter([x], [y], marker='o', s=400, c='firebrick', alpha=0.8)
    if i == 0:
        ax.set_ylabel('Duration [min]', fontsize=24)
    ax.set_xlabel('Rainfall Area [km$^2$]', fontsize=24)
ary=im.get_array()/im.get_array().sum() * 100
im.set_array = ary
cbar = fig.colorbar(im, cax=cbar_ax, orientation='horizontal')
tmp = cbar.ax.tick_params(labelsize=24)
tmp = cbar.set_label('Density [ ]',size=24)
#cbar.set_ticks(np.linspace(vmin, vmax, nticks).round(2))
#cbar.set_ticklabels(np.linspace(vmin, vmax, nticks).round(2))
#print('Medians:')
#medians
#print(' '.join(['%s: %s'%(k, v) for (k,v) in medians.items()]))
#plt.show()

<matplotlib.figure.Figure at 0x7f69f09e0e80>

colm = matplotlib_to_plotly(CubeYF_20.get_mpl_colormap(N=8, gamma=2.0),8, rgb=False)

3. Storm tracks above the 80th percentiles

• marks start of track, colors for different ensemble member

#Plot the tracks
mpld3.enable_notebook()
fig = plt.figure(figsize=(14,5.5))
fig.subplots_adjust(right=0.94, bottom=0.45, top=0.95,left=0.01, hspace=0, wspace=0)
#cbar_ax = fig.add_axes([0.12, 0.17, 0.74, 0.02])
o = namedtuple('Sim', 'dataset percentiles')
tmp_obs = o({'obs': CPOL_t.dataset['obs']}, {'obs': CPOL_t.percentiles['obs']})

with nc(CPOLF) as fnc:
    lon=fnc.variables['longitude'][:]
    lat=fnc.variables['latitude'][:]
tp = 'mean'
num=80
titels = ['CPOL', 'UM 1.33km', 'UM 0.44km']
m = None
for i, data in enumerate(((tmp_obs, storm_obs), (UM133_t, storm_UM133), (UM044_t, storm_UM044))):
    tracks, stormId = data
    ax = fig.add_subplot(1,3,i+1)
    #ax2 = fig.add_subplot(2,3,i+4)
    ax.set_title(titels[i], fontsize=18)
    for ii, tr in enumerate(tracks.dataset.keys()):
        if ii == 0:
            draw_map = None
            m2 = None
        else:
            draw_map = m
        Id = [stormId.Sim[tr]]
        perc = tracks.percentiles[tr][tp][num]
        ax, m, im = plot_traj(tracks.dataset[tr], lon, lat, ax=ax, mintrace=2, thresh=('mean', perc), 
                  color=colm[ii], create_map=draw_map, basemap_res='f', lw=0.5, size=20, particles=None)
        #ax2, m2, im = plot_traj(tracks.dataset[tr], lon, lat, ax=ax2, mintrace=2, thresh=('mean', perc),
        #          color=colm[ii], create_map=m2, basemap_res='f', lw=0.5, size=20, particles=Id)
        #break
#plt.show()

4. Classification of strom area, rain-rate and duration by rain-rate quintiles

mpld3.disable_notebook()
variables=dict(dur=('Duration [min]',300), avg_area=('Area [km$^2$]', 100*2.5**2), avg_mean=('Rain-rate [mm/h]',12))
fig = plt.figure()
fig.subplots_adjust(right=0.94, bottom=0.025, top=0.6,left=0.01, hspace=0, wspace=0)
i = 0
for y, prop in variables.items():
    label, ylim = prop
    npl = len(list(variables.keys()))
    i += 1
    ax = fig.add_subplot(npl,1,i)
    ax = sns.boxplot(x="quant", y=y, hue="run", data=df_stack, palette="muted", ax=ax)
    #ax = sns.stripplot(x="quant", y=y, hue="run", data=df_stack, jitter=True, palette="Set2", dodge=True)
    ax.legend(loc=0, fontsize=24)
    ax.tick_params(labelsize=24)
    ax.set_xlim(0.5,5.5)
    ax.set_ylim(0,ylim)
    ax.yaxis.set_ticks(ax.yaxis.get_ticklocs()[1:])
    ax.set_xlabel('Quintiles', fontsize=26)
    ax.set_ylabel(label, fontsize=26)
    #break
#fig.savefig(os.path.join(os.environ['HOME'], 'Todds_Rainfall_2.pdf'), bbox_inches='tight', format='pdf', dpi=300)

5. Stormtrack Percentiles vs Rain-rate

P = list(np.arange(0,99))+[99, 99.9, 99.99, 99.999, 100]

PERC = pd.DataFrame({'Obs':np.percentile(CPOL_pdf2['avg_mean'].values, P), 
                     'UM 1.33km': np.percentile(UM133_pdf['avg_mean'].values, P),
                      'UM 0.44km': np.percentile(UM044_pdf['avg_mean'].values, P)},index=P)

from mpl_toolkits.axes_grid.inset_locator import inset_axes
#Plot the percentages
fig=plt.figure()
ax = fig.add_subplot(111)
plt.subplots_adjust(bottom=0.05, right=0.9, top=0.4)
ax.plot(PERC.index, PERC['Obs'].values, color='lightseagreen', linestyle='-', label='CPOL',lw=3)
ax.plot(PERC.index, PERC['UM 1.33km'].values, color='fuchsia', linestyle='--', label='UM 1.33km',lw=3)
ax.plot(PERC.index, PERC['UM 0.44km'].values, color='fuchsia', linestyle='-', label='UM 0.44km', lw=3)
ax.fill_between(PERC.index, member.min(axis=0)-0.25, member.max(axis=0)+0.25, color='cornflowerblue', alpha=0.2)
ax.plot(PERC.index, member.mean(axis=0), color='lightseagreen', linestyle='--', label='Bootstrap',lw=3)
ax.set_xlim(1,100)
#ax.set_ylim(10,100)
#ax.set_yscale('log')
#ax.set_xscale('log')
ax.set_xlabel('Percentile [\%]', fontsize=26)
ax.set_ylabel('Rain-rate [mm/h]', fontsize=26)
ax.legend(loc=0, fontsize=24)
ax.tick_params(labelsize=24)
ax.grid(color='r', linestyle='-', linewidth=0)

#Bootstrapping stuff:

8. Cold-Pools

Cold-Pool tracking using density potential temp. pertubation fields

embed_vid(os.path.join(dest_dir, 'ColdPool-Ens-2.mp4'))

mpld3.disable_notebook()
variables=dict(cp_area=('Cold-Pool Area [km$^2$]', 220), cp_mean=('Cold-Pool strength [K]', 2), 
               cp_dur=('Cold-Pool Lifetime [min]', 120))
fig = plt.figure()
fig.subplots_adjust(right=0.94, bottom=0.025, top=0.6,left=0.01, hspace=0, wspace=0)
i = 0
for y, prop in variables.items():
    label, ylim = prop
    npl = len(list(variables.keys()))
    i += 1
    ax = fig.add_subplot(npl,1,i)
    ax = sns.boxplot(x="quant", y=y, hue="run", data=df_stack, palette="muted", ax=ax)
    #ax = sns.stripplot(x="quant", y=y, hue="run", data=df_stack, jitter=True, palette="Set2", dodge=True)
    if i == 1:
        ax.legend(loc=0, fontsize=24)
    else:
        ax.legend(loc=0, fontsize=0.001)
    ax.tick_params(labelsize=24)
    ax.set_xlim(0.5,5.5)
    if i == 3:
        ax.set_ylim(60,ylim)
    else:
        ax.set_ylim(0, ylim)
    ax.yaxis.set_ticks(ax.yaxis.get_ticklocs()[1:])
    ax.set_xlabel('Quintiles', fontsize=26)
    ax.set_ylabel(label, fontsize=26)
    #break
#fig.savefig(os.path.join(os.environ['HOME'], 'Todds_Rainfall_2.pdf'), bbox_inches='tight', format='pdf', dpi=300)

Fluxes

• Distributions of some variables

mpld3.disable_notebook()
variables=dict(omega=('Vert. motion', (0, 6)), bflux=('Bouyancy Flux', (-3,30)), mfluxdiv=('Moist. Flx Conv',(-5,6)),
               momflux=('Momentum Flux', (-2, 2)), moistflux=('Moisture Flux', (-0.01, 0.04)))
               #cloud_w=('Cloud Water', (0,0.005)))
fig = plt.figure()
fig.subplots_adjust(right=0.94, bottom=0.025, top=0.6,left=0.01, hspace=0, wspace=0)
i = 0
for y, prop in variables.items():
    label, ylim = prop
    npl = len(list(variables.keys()))
    i += 1
    ax = fig.add_subplot(npl,1,i)
    ax = sns.boxplot(x="quant", y=y, hue="run", data=storm_prop, palette="muted", ax=ax)
    #ax = sns.stripplot(x="quant", y=y, hue="run", data=df_stack, jitter=True, palette="Set2", dodge=True)
    if i == 1:
        ax.legend(loc=2, fontsize=24)
    else:
        ax.legend(fontsize=1)
    ax.tick_params(labelsize=24)
    ax.set_xlim(0.5,5.5)
    ax.set_ylim(*ylim)
    ax.yaxis.set_ticks(ax.yaxis.get_ticklocs()[1:])
    ax.set_xlabel('Quintiles', fontsize=26)
    ax.set_ylabel(label, fontsize=26)

• Some vertical profiles

fig = plt.figure()
from matplotlib.ticker import Formatter

from scipy import interpolate
variables=dict( momflux=('Momentum Flux', (-1.5, 1.3)),
               mfluxdiv =('Moist. Flx Conv', (-0.5e-3,1.3e-3)),
               mflux=('Moisture Flux', (-1.2e-4, 1.6e-3)), omega=(('Vert. Motion'), (-9.5e-2, 8.5e-1)))
i = 0
fontsize=24
y = np.linspace(P[-1], P[0], 40)[::-1]
for x, prop in variables.items():
    label, xlim = prop
    npl = len(list(variables.keys()))
   
    for ii in range(1, 6):
        i += 1
        ax = fig.add_subplot(npl,5,i)
        for ff, fluxes in enumerate((fluxes133, fluxes044)):           
            flux = fluxes[x][ii]
            f = interpolate.interp1d(P, flux, kind='slinear')
            #xx = f(y)
            y = P
            xx = flux
            ax.plot(xx, y, label=('UM 1.33km', 'UM 0.44km')[ff])
            ax.set_ylim(max(y), min(y))
            ax.set_xlim(*xlim)
            if ii == 1:
                ax.set_ylabel('Pressure [hPa]', fontsize=fontsize-2)
            else:
                ax.yaxis.set_ticklabels([])   
            if i == 1:
                 ax.legend(loc=1, fontsize=fontsize)
            else:
                ax.legend(loc=1, fontsize=0.001)
            ax.set_xlabel(label, fontsize=fontsize-2)
            ax.tick_params(labelsize=fontsize-2)
            ax.ticklabel_format(axis='x', style='sci')
            if i < 6:
                ax.set_title('Qantile  %i'%ii, fontsize=fontsize)
            #ax.xaxis.set_major_formatter('%.1E')
            ax.ticklabel_format(style='sci', axis='x') 
            ax.xaxis.major.formatter.set_powerlimits((0,0))
fig.subplots_adjust(right=0.98, bottom=0.25, top=0.99, left=0.01, hspace=0.15, wspace=0.04)